Logic and memory in single unit enables Toshiba to reduce FPGA die size by half

Toshiba has developed space-saving and low-cost nonvolatile memory based circuit technology for field-programmable gate arrays (FPGAs). Toshiba disclosed the details on 14 June at the "2016 Symposium on VLSI Technology", an international conference on semiconductor devices in Honolulu, Hawaii. With the traditional CPU lacking in per watt performance for handling high-definition media stream processing or any such continuous stream of complex data, custom processors are evolving faster either as custom ASIC or through FPGAs. The advantage of FPGAs over ASIC is the logic circuits can be altered without changing the chip. FPGA vendors are offering SOC chips integrated with ARM Cortex processor cores and FPGA fabric, so that some new embedded system applications can leverage both the traditional CPU as well as custom processors wired inside the FPGA fabric. Modern artificial intelligence (AI) neuron chips use lot of custom processing hardware. To give examples of that trend, the leading FPGA vendor Xilinx is offering Zynq UltraScale+ MPSoC family with the option of having quad ARM Cortex-A53s, dual Cortex-R5s, a graphics processing unit, and a video codec unit for applications such as motor control, sensor fusion, medical, and handheld radios. Intel owned Altera and Microsemi are also offering such chips and are popular. Data centres are extensively using FPGA for cust...